Does any solution exist for RTC with following consideration?

  • Does not have to be battery backed. The time can be set using a network clock at the OS (Linux) startup.
  • The time needs to be accessible at high memory access speeds (I2C may be slow). Possibly mmap'd to user space for reading the time at least.
  • Provides millisecond granularity.
  • Does ARM processor itself provide any such capability? Perhaps a count down register that can be used to emulate real time keeping?
  • The normal system clock is inevitably hardware based (hopefully someone can provide details in case you are dubious), and generally speaking the OS should provide access to a "steady clock" programmatically, that is, one which will not vary during the lifetime of a process. It will have micro- or milli- second precision. – goldilocks Feb 7 '17 at 19:38
  • Thanks for your response. Actually I do see significant clock drift on Orange Pi (stackoverflow.com/questions/41575617/…), and have some details. Although I believe I have a workaround I am beginning to think I need to have a better solution. Also per tldp.org/HOWTO/Clock-2.html the two clocks can drift. – Cap Feb 7 '17 at 20:15
  • Two PC clocks can drift, and they are hardware. That's the point of NTP -- much more expensive hardware somewhere far away that doesn't drift. – goldilocks Feb 7 '17 at 20:17
  • Possibly you want a few of these. – goldilocks Feb 7 '17 at 20:21
  • 1
    Thanks again. My concern is that a few times during Radiotap captures I noted the clock to drift significantly from the NTP time over a short duration (minutes). For the requirement a drift of a few seconds per day is also tolerable, and a Cesium clock would be an overkill. Eventually if no alternative is available I may work around by using an I2C RTC, but would wait for any other suggestions. – Cap Feb 7 '17 at 20:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.